Local environmental quality and life-satisfaction in Germany

Local environmental quality and life-satisfaction in Germany

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 08 ) 78 7 –7 97 a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m w w w. e l s e v i e r. ...

316KB Sizes 0 Downloads 76 Views

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 08 ) 78 7 –7 97

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

w w w. e l s e v i e r. c o m / l o c a t e / e c o l e c o n

ANALYSIS

Local environmental quality and life-satisfaction in Germany Katrin Rehdanz a,⁎, David Maddison b a

Research Unit Sustainability and Global Change, Hamburg University and Centre for Marine and Atmospheric Science, Bundesstrasse 55, 20146 Hamburg, Germany b Department of Economics, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK

AR TIC LE I N FO

ABS TR ACT

Article history:

Hitherto the task of valuing differences in environmental quality arising from air pollution

Received 23 September 2006

and noise nuisance has been carried out mainly by using the hedonic price technique. This

Received in revised form

paper proposes a different approach to derive information on individual preferences for

15 April 2007

local environmental quality. It analyses data drawn from the German socio-economic panel

Accepted 30 April 2007

in an attempt to explain differences in self-reported levels of well-being in terms of

Available online 7 June 2007

environmental quality. Mindful of existing research a large number of other explanatory variables are included to control for socio-demographic differences, economic

Keywords:

circumstances as well as neighbourhood characteristics. Differences in local air quality

Air pollution

and noise levels are measured by how much an individual feels affected by air pollution or

Environmental quality

noise exposure in their residential area. The evidence suggests that even when controlling

Germany

for a range of other factors higher local air pollution and noise levels significantly diminish

Life-satisfaction

subjective well-being. But interestingly differences in perceived air and noise pollution are

Noise exposure

not capitalised into differences in house prices.

Well-being

1.

Introduction

In most developed countries standards for air pollution and noise exposure are an important part of environmental policy to improve local environmental quality. Often these standards are based on expert judgements and do not take into account peoples preferences. Although improvements in air quality or reductions in transportation noise are beneficial, they are nevertheless costly. As this money could be spent elsewhere in an economy, information on the social benefits is needed prior to implementing such standards. Two valuation techniques have typically been used to derive the welfare benefits from reduced air pollution or noise nuisance. The hedonic price approach relies on data from the

© 2007 Elsevier B.V. All rights reserved.

housing market whilst the contingent valuation method derives values by asking people directly about their willingness to pay. The contingent valuation method is often criticized because of the difficulties related to the construction of a market and its numerous potential biases (see Mitchell and Carson, 1989 or Bateman et al., 2002, for a more recent overview of the method). The majority of studies apply the hedonic price approach to measure the benefits of air quality improvements or to value noise (see Freeman, 2003, for an overview of the method). All hedonic studies assume that the ability of households to relocate eliminates the net benefits of different locations whilst simultaneously giving rise to compensating house price

⁎ Corresponding author. Research Unit Sustainability and Global Change, Hamburg University and ZMAW, Bundesstrasse 55 (Pavillon), 20146 Hamburg, Germany. Tel.: +49 40 428387047; fax: +49 40 428387009. E-mail address: [email protected] (K. Rehdanz). 0921-8009/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolecon.2007.04.016

788

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 8 ) 7 8 7 –7 97

and in some cases, wage rate differentials.1 This requires the existence of perfect information, zero moving and transaction costs, as well as perfectly flexible prices. That these assumptions are more realistic for some countries than others is obvious. Perhaps it is not therefore very surprising that most of what is known about societal willingness to pay for improvements in air quality or reductions in noise nuisance comes from hedonic studies carried out in the United States or the United Kingdom. For Germany only one early hedonic study exists.2 Holm-Müller et al. (1991) analyse data on (selfreported levels of) noise exposure for three cities in Germany (Bielefeld, Bremen and Wuppertal) and find no significant effect of road traffic noise on property prices. For overviews of the empirical literature see Smith and Huang (1995) for air pollution studies and Nelson (2004) for noise pollution studies linked to airports. This paper proposes a different approach to valuing air and noise pollution using data on subjective well-being (SWB). More specifically, the paper considers the link between perceived levels of noise and air pollution and self-reported happiness using individual-level data. We further test the stability of these regressions over both time and space and in addition test the extent to which any welfare impacts caused by noise nuisance and air pollution are also capitalised into hedonic house price regressions. The remainder of the paper is structured as follows. Section 2 presents an overview of the economic research into the determinants of happiness focussing in particular on those few studies that have attempted to link happiness with environmental quality. Section 3 presents the analytical framework underlying the alternative approaches. Section 4 describes data sources and Section 5 the empirical implementation. Section 6 discusses the results and the final section concludes.

2.

Literature review

One strand of economics is based on analysing people's decisions revealed through their market behaviour. Investigating, e.g., people's choices on the property market, should reveal their preferences for locations including environmental amenities as well as property characteristics. The hedonic price method takes advantage of this kind of information. However, an alternative is to focus on “experienced” utility, stressing the pleasure of consumption, as provided by survey measures of happiness or life-satisfaction (see for example Kahnemann et al., 1997; or more recently Kahnemann and Sudgen, 2005). 1

It is difficult to see how highly localised environmental goods such as air quality and noise nuisance can give rise to compensating wage differentials. For a theoretical model in which environmental quality can be capitalised into both house prices and wage rates see Roback (1982). 2 Three contingent valuation studies exist; Schulz (1985), HolmMüller et al. (1991) and Weinberger et al., 1991). Schulz (1985) values improvements of air quality in Berlin. Weinberger et al. (1991) analyse noise exposure and Holm-Müller et al. (1991) study a variety of issues including air and water quality, noise exposure and amenity areas.

Research into the determinants of happiness or lifesatisfaction was for a long time solely the domain of psychologists and sociologists. In recent years however, it has become increasingly accepted into welfare economics. Today there is the general belief that data on subjective wellbeing are valid and can be used for formal analyses (Di Tella et al., 2003). Empirical work has furthermore shown that happiness is not a purely personal issue, but that economic conditions like income, unemployment and inflation also have a strong impact on people's subjective well-being (see e.g. Clark and Oswald, 1994; Di Tella et al., 2001; Easterlin, 2001). See Frey and Stutzer (2002) for a review of the literature. A handful of studies also attempt to explain differences in SWB as a function of ambient environmental quality. Air pollution was found to reduce happiness in a study using country level data (Welsch, 2002, 2006). Rehdanz and Maddison (2005) also found climate variables to have a highly significant effect on average self-reported levels of happiness. Frijters and Van Praag (1998) estimate the effects of climate on both welfare and well-being in Russia finding climate to be an important determinant of households' standard of living. Van Praag and Baarsma (2005) studied the area around Amsterdam Airport and found exposure to aircraft noise to reduce SWB to a significant extent. Presenting a slightly different type of analysis, Ferrer-i-Carbonell and Gowdy (2007) analyse the relationship between subjective measures of well-being and environmental attitudes.

3.

Analytical framework

In the hedonic framework house prices and the levels of environmental attributes are endogenous to the household in the sense that the household chooses where to locate. The household maximises utility, which is a function of the quantity of some composite good and the level of the environmental attribute. This maximisation process is subject to a budget constraint linking household income to expenditure on the composite commodity. For an individual who occupies the house j the utility is given by: maxU ¼ UðX; Q j ; Sj ; Nj Þ s:t: I ¼ X þ R j

ð1Þ

where: X Q S N I R

is the numeraire good traded on the market, is a vector of location-specific environmental amenities, is a vector of structural characteristics of the house, is a vector of neighbourhood characteristics, is income, is rental payments.

The expenditure on housing which is itself a function of the level of the environmental attribute is given by: R j ¼ RðSj ; Nj ; Q j Þ:

ð2Þ

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 08 ) 78 7 –7 97

The first-order condition for the choice of environmental quality q is given by: AU=Aq AR j ¼ : AU=AX Aq

ð3Þ

The household's marginal willingness to pay is revealed by the household's position on the implicit price schedule, which plots the gradient of the hedonic price function with respect to the level of the amenity (Rosen, 1974). In the hedonic model calculating the value of non-marginal changes is more involved since the inverse demand function for the environmental amenity is not identified unless the households are observed under different supply and demand conditions. The hedonic approach rests on the assumption that markets are in equilibrium and individuals can freely move between locations without transaction costs. This implies that perfectly mobile individuals would settle where they could maximise their net benefits. But in Germany, laws relating to rented property mean that Germans tend to be less mobile compared to other nations. About 60% of all households live in rented accommodation (Hoffmann and Kurz, 2002). Until recently, tenants and landlords had to comply with periods of notice to up to twelve months. As a result, the monetary costs of moving were considerable. Since the end of 2001 the legal period of notice in new contracts is reduced to three months for tenants, regardless of the duration of the contract, to better account for the changed situation in the labour market where flexibility is of key importance (Grundmann, 2001). Mobility of tenants is further restricted as sitting tenants benefit from length-of-stay discounts which increase relative to the length of the tenancy. There are also limits on the ability of landlords to increase rents (no more than 30% in any threeyear period). The most substantial increases always take place when tenants move (Hoffmann and Kurz, 2002). This causes a lock-in effect: even if the old flat does not suit the needs of the household any more old contracts are not available to potential new tenants and new contracts might not be attractive to sitting tenants. In addition the non-monetary psychic costs of leaving the neighbourhood and the neighbours one is accustomed might prevent moving. For these reasons implicit prices might not fully reflect the household's marginal willingness to pay for environmental attributes. In the second approach presented in this paper the indirect utility function is estimated directly and all prices and levels of environmental attributes are taken as exogenous. For an individual living in house j the indirect utility is given by: U ¼ UðP; Q j ; Sj ; Nj ; I; RÞ

ð4Þ

where: P Q S N I R

is a vector of personal characteristics, is a vector of location-specific environmental amenities, is a vector of structural characteristics of the house, is a vector of neighbourhood characteristics, is income, is rental payments.

789

In this framework marginal willingness to pay for the environmental attribute is given by the derivative of the indirect utility function with respect to the level of the environmental attribute divided by the derivative of the indirect utility function with respect to income. And unlike with the hedonic approach calculating the value of nonmarginal changes in level of the environmental amenity is straightforward. How are the two approaches described above related? If the hedonic approach fails to reflect fully environmental quality differences, the derived implicit price underestimates the marginal willingness to pay. In their paper, Van Praag and Baarsma (2005) refer to a residual shadow cost. They argue that the total shadow cost is the sum of the hedonic price differential and the residual shadow cost. Our analysis follows theirs in firstly estimating to what extent residual shadow costs for changes in air and noise pollution are reflected in the indirect utility function, then estimating the implicit price using the hedonic approach.

4.

Empirical analysis

Most of the data used in this study is taken from the German socio-economic panel (SOEP) survey. The SOEP is based on a set of pre-tested questionnaires for both households and individuals. Since 1984 the survey has provided annual information on housing, as well as on the occupation, employment history and earnings of individuals. It was extended to include former East Germany in 1990. In addition to a stable set of core questions, each year the survey focuses on a special topic. In 1994, 1999, and 2004 the questionnaire asked for details of housing and neighbourhood characteristics including information on how strongly the respondent feels affected by air and noise pollution in their place of residence.3 One feature of the survey is that the head of the household only answers the questions on housing and neighbourhood characteristics. The head of the household is defined as the person who knows best about the general conditions under which the household operates. However, the questions on air pollution and noise exposure are directed to the individual and should not be answered on behalf of other household members. Tables A1 and A2 in Appendix A provide information on the distribution as well as the exact wording of the two questions. In order to take advantage of this information the analysis relies exclusively on the surveys of 1994, 1999 and 2004 and on information the household head provided and excludes information provided by other household members. Taken together, our dataset contains a total number of about 23,000 observations. Over the years some households dropped out and were replaced. The size of the survey is also increasing

3

The issue that the analysis is based on a subjective measures rather than objective measures of air and noise pollution is discussed in more detail in Section 6.

790

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 8 ) 7 8 7 –7 97

Table 1 – Definition of variables included in the two regression models Variable AIR

NOISE

AGE MALE INCOME EDUCATION UNEMPL CHILDREN DISABLED HEALTH TOGETHER RENT CONDITION

WINDOW AREA

TYPE

BUILT

SQUARE CITY

TRANSPORT

OWNER SOCIAL NOHEATING GARDEN BALCONY MOVEIN GGK

STATE

Definition Adversely affected by level of air pollution (1 = not at all, 2 = slightly, 3 = bearable, 4 = strongly, 5 = very strongly) Adversely affected by noise pollution (1 = not at all, 2 = slightly, 3 = bearable, 4 = strongly, 5 = very strongly) Age of the individual Unity if the individual is male, zero otherwise Household net income adjusted for inflation (in EURO) Institutional years necessary to receive the current degree of education Unity if individual is registered unemployed, zero otherwise Number of children living in household Unity if individual is disabled, zero otherwise Current health status (1=very good, 2=good, 3=satisfactory, 4=poor, 5=bad) Unity if individual lives together with a partner, zero otherwise Monthly housing expenses (in EURO) Condition of property (1=good, 2=needs renovation, 3=needs complete renovation, 4=ready for demolition) Unity if the property got new windows last year, zero otherwise Kind of residential area (1=mostly old houses (built before the war), 2=mostly newer houses, 3=residential and commercial area with flats, houses, shops and businesses, 4=commercial area (shops, banks, offices) with few flats, 5=industrial area with few flats, 6=else) Building type (1=agricultural building, 2=single or double house, 3=terrace house, 4=flat in building with 3 to 4 flats, 5=flat in building with 5 to 8 flats, 6=flat in building with 9 or more flats, 7=flat in high rise building or 8=else) Vintage class (1=before 1919, 2=1919–1948, 3=1949–1971, 4=1972–1980, 5=1981–1990, 6=1991 or later) Size of property in square meters Distance to centre of the nearest city (1=property is in city centre, 2=less than 10 km, 3=10–25 km, 4=25–40 km, 5=40–60 km, 6=60 km or more) Walking distance to public transport stop (1=under 10 min, 2=10–20 min, 3=more than 20 min, 4=not in walking distance) Unity if property is owned, zero otherwise Unity if the property is a council house, zero otherwise Unity if property has no central heating, zero otherwise Unity if the property has a garden, zero otherwise Unity if the property has a balcony, zero otherwise Year the household moved in Community size (1=less than 2000, 2=2000– 20,000, 3=20,000–100,000, 4=100,000–500,000, 5=more than 500,000 inhabitants) Federal State (1=Schleswig-Holstein, 2=Hamburg, 3=Lower Saxony, 4=Bremen, 5=North RhineWestphalia, 6=Hesse, 7=Rhineland-Palatinate and Saarland, 8=Baden-Wuerttemberg, 9=Bavaria, 10=Berlin, 11=Brandenburg, 12=Mecklenburg Western-Pommerania, 13=Saxony, 14=Saxony-Anhalt, 15=Thuringia)

Table 1 (continued) Variable EAST

UNEMPL _STATE GDP _STATE 2004 1999

Definition Unity if Federal State belongs to Eastern Germany (Berlin is matched to the West German sample), zero otherwise Rate of unemployment within a Federal State Average GDP per capita per Federal State Unity if the observations are drawn from the 2004 survey, zero otherwise Unity if the observations are drawn from the 1999 survey, zero otherwise

Source: German socio-economic panel, Statistisches Bundesamt (2006a, 2006b), Bundesagentur für Arbeit (2006).

over time. In 1994 information on 6131 participants is available for analysis compared to 6643 in 1999 and 10,415 in 2004. Turning to the analysis of SWB first, SWB is measured on an integer scale of 0–10 with an average SWB of 6.79 for our sample (see Table A3 in Appendix A for more detailed information). A feature of the survey is that people living in former East Germany tend to be less happy compared to those in West Germany. One explanation might be the lower income levels and the higher rate of unemployment in the East. Therefore, a dummy variable separating Germany into East and West is included. Other explanatory variables, generally found significant in models explaining differences in SWB, are individual characteristics such as age, gender, employment status, education level, income, number of children, physical condition or living together with a partner. Note that the income data was adjusted for inflation using information on the consumer price indices (Statistisches Bundesamt, 2006a). In addition to the above, we include characteristics of the property the individual is living in (age of the building, type of building, condition of the property, size of the property) as well as neighbourhood characteristics to identify possible factors influencing air pollution or noise exposure. Dummy variables indicate whether the individual is living, for example, in a predominantly residential area, industrial or commercial area. Controls are also included for the size of the town or city as well as the distance to the nearest large city or the closest transport link. To capture regional differences on a more disaggregated level dummy variables indicate in which of the Federal State the individual is living. Further dummy variables indicate whether observations are drawn from the 1994, 1999 or the 2004 survey along with regional income per capita and the unemployment rate. Turning to the hedonic regression, the logarithm of monthly rental costs per square meter was regressed on a number of environmental characteristics including air pollution and noise exposure as well as neighbourhood characteristics and structural attributes of the property. Note that for owners the survey provides self-reported imputed rents rather than actual rents. For 2004 no information on rents for owners is provided. The hedonic analysis is, therefore, restricted to data from 1994 to

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 08 ) 78 7 –7 97

1999. Also, we excluded from our analysis households living in residential homes, student halls and hostels. The same set of variables describing the environmental, property and neighbourhood characteristics are used in both the hedonic equations as well as in the equations explaining differences in SWB. In addition, the hedonic model controls for the year the household moved in, whether the house has heating, a garden, a balcony, is owner occupied or is a council house. All those variables are likely to have a significant influence on the rental rate. In line with the demands of theory individuals' socio-economic characteristics are excluded from the hedonic regression. In order to account for the possible correlation of residuals when observations are taken from the same individual over the different years, the standard errors of both models were adjusted for clustering. The effect is to increase the standard errors of the parameter coefficients. This procedure also leads to robust variance estimates in the face of heteroscedasticity. A list of variables contained in the dataset is presented in Table 1.

5.

Results

5.1.

The SWB model

The SWB data are analysed using an ordered probit model and the results of different model specifications presented in Table 2. These indicate that when all the data is pooled together there is a clearly discernible effect of both perceived noise and air quality on SWB. Note that the inclusion of air and noise quality as categorical rather than continuous variables does not result in a statistically significant increase in fit.4 Analysing the data from each year separately does not yield markedly different results. Dividing the data according to whether households are located in former East or West Germany does not alter the regression coefficients either although the statistical significance of the results for former East Germany is much diminished. Therefore, we formally test whether the coefficients describing the impact of higher air pollution or noise exposure are homogenous across space and time. Results of these parameter homogeneity tests are presented in Table A4 in Appendix A. The null hypothesis of parameter homogeneity is not rejected for air pollution. The variance-weighted estimates are highly significant across space and time. For noise exposure the null hypothesis of parameter homogeneity over time is rejected on the fivepercent level of significant. However, when steps are taken to test parameter homogeneity over time for smaller regions (East and West separately), the test is not rejected, but the coefficient is significant for Western Germany only (results are not shown). The German socio-economic panel shows that the number of individuals believing themselves to be strongly or very strongly affected by air pollution has decreased over time (compare Tables A1 and A2 in Appendix A). This is consistent The χ2(3) statistic of the Wald-test ranges from 2.75 (p = 0.4311) to 6.83 (p = 0.0775) for air pollution and from 0.32 (p = 0.9553) to 4.56 (p = 0.2066) for noise.

791

with the reductions in air pollution that have occurred over that time period. The same set of results can be observed for noise nuisance. But the fact that the number of individuals falling into different categories has changed should not itself alter the coefficient measuring the disutility associated with changes in air pollution. Although on average pollution levels have for most pollutants been declining over time for as long as we are able to find sufficient numbers of individuals who are strongly or very strongly affected by air pollution we are able to estimate the disutility of air and noise pollution. Since 1974 air pollution became subject to more and more regulations in Germany. Whilst at the beginning mostly new plants were regulated, by the mid 1980s all power plants and larger industrial plants were regulated partly in order to meet the requirements of the first sulphur protocol. In the 1990s, after reunification, industrial and power plants in former East were either shut down or completely modernized. Improvements in energy efficiency prompted by concerns over climate change helped keep air pollution at low levels (Umweltbundesamt, 2006). In recent years however measurements of ambient PM10 and PM2.5 have shown an increasing tendency to exceed statutory thresholds. Epidemiological studies indicate that current ambient levels of particulates are associated with a statistically significant change in the likelihood of experiencing a range of adverse health effects and reduced life expectancy (Umweltbundesamt, 2005). Consistent with earlier analyses we find a U-shaped relationship with age and a diminishing marginal utility of money. Unemployed, sick and single individuals are significantly less happy than their counterparts. Individuals living in poor housing or paying higher rents are also less happy. Although in Table 2 above the estimated coefficients of the ordered probit models have no meaningful interpretation, the impact of changes can be computed nevertheless.5 Taking advantage of this information it is possible to compute the amount of money needed to compensate for a reduction or the willingness to pay for an improvement in environmental quality to attain the same level of utility (SWB). For the first model specification (denoted ALL) about EUR 690 per household and month would be needed to compensate for a reduction from one category to the next lower; for an increase in noise exposure about EUR 390 are needed. Although possible to compute, the significance of those numbers is limited as moving from one category of impairment to the next higher might be particularly difficult (as well as expensive) to achieve; especially for the top categories. More than 70% of all respondents feel only slightly or not at all affected by air or noise pollution. They might not notice improvements or might value them only little. Therefore, it might be more important to improve the quality for those in the last three categories. This might be easier to achieve and also less costly. In our model, improving the quality of air and noise for those in the bottom two categories (“very strongly affected” and “strongly affected”) to medium quality (“bear-

4

5 How to calculate the impact of changes on explanatory variables on SWB is explained in more detail in Appendix B.

792

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 8 ) 7 8 7 –7 97

Table 2 – Subjective well-being, noise and air quality

AIR NOISE AGE AGE2 MALE INCOME INCOME2 EDUCATION UNEMPL CHILDREN CHILDREN2 DISABLED HEALTH TOGETHER RENT SQUARE CONDITION WINDOW AREA1 AREA2 AREA3 AREA4 AREA5 TYPE1 TYPE2 TYPE3 TYPE4 TYPE5 TYPE6 TYPE7 BUILT CITY TRANSPORT GGK1 GGK2 GGK3 GGK4 UNEMPL_STATE GDP_STATE STATE1 STATE2 STATE3 STATE4 STATE5 STATE6 STATE7 STATE8 STATE9 STATE10 STATE11 STATE12 STATE13 STATE14 STATE15 EAST 1999 1994 cut1 b cut2 cut3 cut4 cut5 cut6

All a

West a

East a

2004

1999

1994

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

− 6.11E−02⁎⁎⁎ − 3.42E−02⁎⁎⁎ − 2.57E−02⁎⁎⁎ 3.45E− 04⁎⁎⁎ − 3.92E−02⁎⁎ 9.28E− 05⁎⁎⁎ − 1.05E−09⁎⁎⁎

−5.07E−02⁎⁎⁎ −4.82E−02⁎⁎⁎ −2.01E−02⁎⁎⁎ 2.84E−04⁎⁎⁎ −5.08E−02⁎⁎⁎ 7.66E−05⁎⁎⁎ −8.46E−10⁎⁎⁎

−8.52E−02⁎⁎⁎

−5.21E− 02⁎⁎⁎ −6.53E− 02⁎⁎⁎ −3.29E− 02⁎⁎⁎ 4.11E−04⁎⁎⁎

−6.49E− 02⁎⁎⁎

−7.41E− 02⁎⁎⁎ −1.16E− 02 −1.87E− 02⁎⁎⁎ 2.84E−04⁎⁎⁎

− 2.34E−03 − 5.06E−01⁎⁎⁎ − 9.01E−02⁎⁎⁎ 1.90E− 02⁎⁎⁎

−3.47E−03 −5.07E−01⁎⁎⁎ −9.09E−02⁎⁎⁎ 1.86E−02⁎⁎⁎

−7.43E−04 −4.40E−01⁎⁎⁎ −1.24E−01⁎⁎⁎ 2.99E−02⁎⁎

− 3.06E−02 − 5.35E−01⁎⁎⁎ 1.88E− 01⁎⁎⁎ − 6.82E−05⁎⁎⁎ 6.62E− 04⁎⁎ − 1.47E−01⁎⁎⁎

−1.94E−02 −5.34E−01⁎⁎⁎ 2.06E−01⁎⁎⁎ −7.21E−05⁎⁎⁎ 7.62E−04⁎⁎⁎ −1.71E−01⁎⁎⁎

−7.29E−02 −5.39E−01⁎⁎⁎ 8.29E−02⁎

− 2.23E−02 5.74E− 02 5.56E− 02 7.32E− 02 1.03E− 01 − 1.76E−01 − 1.65E−02 1.16E− 01 1.12E− 01 7.67E− 02 5.12E− 02 7.28E− 03 6.73E− 02 − 1.60E−02⁎⁎⁎ 1.73E− 04 − 5.99E−02⁎⁎⁎ − 9.07E−02⁎⁎ − 6.16E−02⁎

−8.39E−03 4.83E−02 4.40E−02 6.73E−02 9.47E−02 −2.11E−01 −5.99E−02 1.20E−01 1.25E−01 7.67E−02 5.87E−02 2.47E−02 4.82E−02 −1.81E−02⁎⁎⁎ 2.14E−03 −3.61E−02⁎⁎ −1.53E−01⁎⁎⁎ −6.48E−02⁎ −5.84E−02⁎

− 4.92E−02 − 8.87E−03 8.84E− 03 6.65E− 07 4.18E− 01⁎⁎⁎ 5.25E− 01⁎⁎ 4.35E− 01⁎⁎⁎ 4.99E− 01⁎⁎ 4.43E− 01⁎⁎⁎ 5.36E− 01⁎⁎⁎ 5.06E− 01⁎⁎⁎ 4.25E− 01⁎⁎⁎ 4.53E−01⁎⁎⁎ 2.44E− 01⁎⁎⁎

−2.40E−02 1.61E−02 −1.90E−06 −9.48E−02⁎ 4.75E−02 −9.20E−02⁎⁎ −1.97E−02 −7.73E−02 4.12E−02 (Dropped) −5.80E−02 −3.26E−02 −3.27E−01⁎⁎⁎

1.72E− 01 8.20E− 02⁎ 1.41E− 01⁎⁎⁎ (Dropped) 1.28E− 01⁎⁎⁎ (Dropped) 2.37E− 01⁎⁎⁎ 2.87E− 01⁎⁎⁎

(Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) 2.02E−01⁎⁎⁎ 2.82E−01⁎⁎⁎

− 4.33E+ 00 − 4.09E+ 00 − 3.66E+ 00 − 3.17E+ 00 − 2.81E+ 00 − 2.15E+ 00

−4.68E+ 00 −4.45E+ 00 −4.05E+ 00 −3.58E+ 00 −3.23E+ 00 −2.60E+ 00

4.21E−03 −5.09E−02⁎⁎⁎ 6.12E−04⁎⁎⁎ −2.25E−02 3.26E−04⁎⁎⁎ −2.09E−08⁎⁎⁎

−6.68E−05 1.42E−04 −8.31E−02⁎⁎⁎ −2.62E−02 1.88E−01 2.06E−01 1.99E−01 2.20E−01 (Dropped) (Dropped) 6.05E−02 1.48E−02 5.52E−02 −1.93E−02 −7.83E−02 3.43E−02 −1.83E−02 −4.93E−03 −1.31E−01⁎⁎⁎ −6.07E−02 −1.11E−01 −7.46E−02 −3.27E−02 −3.46E−02 2.92E−05 (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) 1.02E−01 6.40E−02 2.92E−02 2.14E−02 (Dropped) (Dropped) 3.43E−01⁎ 2.75E−01 −5.21E+ 00 −4.92E+ 00 −4.43E+ 00 −3.89E+ 00 −3.51E+ 00 −2.75E+ 00

−2.76E− 02 7.20E−05⁎⁎⁎ −8.05E− 10⁎⁎⁎ 9.88E−03 −5.51E− 01⁎⁎⁎ −1.22E− 01⁎⁎⁎ 3.37E−02⁎⁎ −5.78E− 02 −5.14E− 01⁎⁎⁎ 2.⁎28E−01 −5.02E− 05 1.12E−03 −1.70E− 01 3.41E−02⁎⁎⁎ 1.46E−01 1.25E−01 1.37E−01 3.49E−01 −1.27E− 01 −8.68E− 02 1.05E−01 1.23E−01 9.10E−02 7.63E−02 2.50E−02 −2.84E−03 −9.65E− 03 2.91E−03 −6.36E− 02⁎⁎⁎ −7.71E− 02 −1.04E− 01⁎⁎ −8.91E− 02⁎⁎ −2.92E− 02 −2.48E− 02⁎⁎⁎ 7.30E−06⁎ −7.38E− 03 (Dropped) 5.43E−02 1.49E−01 5.87E−02 −2.30E− 02 (Dropped) −1.26E− 01⁎⁎ −8.91E− 02⁎ 1.10E−01 1.43E−01⁎ (Dropped) 3.92E−02 −9.07E− 03 8.76E−03 (Dropped) (Dropped) (Dropped) −4.95E+ 00 −4.69E+ 00 −4.24E+ 00 −3.73E+ 00 −3.36E+ 00 −2.70E+ 00

−1.37E− 02 −2.86E− 02⁎⁎⁎ 3.72E−04⁎⁎⁎ −9.32E− 02⁎⁎⁎ 2.22E−04⁎⁎⁎ −1.73E− 08⁎⁎⁎ −7.17E− 03 −4.43E− 01⁎⁎⁎ −6.81E− 02⁎⁎

−3.71E− 02 2.02E−04⁎⁎⁎ −1.24E− 08⁎⁎⁎ −1.93E− 02⁎⁎⁎ −4.66E− 01⁎⁎⁎ −8.17E− 02⁎⁎⁎

5.73E−03 −2.31E− 02 −5.57E− 01⁎⁎⁎ 2.07E−01⁎⁎⁎ −8.72E− 05⁎⁎⁎

1.38E−02 2.05E−02 −5.49E− 01⁎⁎⁎ 1.24E−01⁎⁎⁎ −7.09E− 05⁎

2.58E−04 −1.34E− 01⁎⁎⁎ 1.84E−02 1.12E−01 2.03E−01 2.03E−01 3.60E−01 −1.93E− 01 3.29E−01 3.27E−01 2.97E−01 2.14E−01 2.50E−01 1.97E−01 1.70E−01 −3.31E− 02⁎⁎⁎ −6.39E− 03 −6.97E− 02⁎⁎⁎ −1.12E− 01 −3.22E− 02 −5.06E− 02 −3.42E− 02 −3.52E− 02⁎⁎⁎

−7.36E− 06 −1.27E− 01⁎⁎⁎ −9.91E− 02⁎ −1.97E− 02 −7.89E− 02 −2.10E− 02 −2.97E− 01 −1.09E− 01 −5.09E− 02 8.89E−02 7.09E−02 6.72E−02 −2.49E−02 −4.45E−02 1.50E−01 −1.38E− 02 3.82E−03 −4.92E− 02⁎ −7.23E− 02 −3.27E− 02 1.31E−03 4.01E−02 −5.30E− 02⁎⁎⁎

7.21E−06 −1.06E− 01 (Dropped) −8.42E−02 1.95E−01 −9.63E− 02 −8.53E− 02 (Dropped) −3.30E− 01⁎⁎⁎ −2.90E− 01⁎⁎⁎ 1.38E−01 7.48E−02 1.09E−01 2.92E−02 (Dropped) −3.57E−02 (Dropped) (Dropped) (Dropped) −5.16E+ 00 −4.92E+ 00 −4.51E+ 00 −4.04E+ 00 −3.69E+ 00 −3.03E+ 00

−2.53E− 06 −4.68E− 02 (Dropped) 1.99E−02 1.72E−01 2.43E−03 4.55E−02 (Dropped) −1.93E− 01⁎⁎⁎ −1.65E− 01⁎⁎ −6.01E−02 −3.46E− 02 2.08E−02 −1.76E− 02 (Dropped) −2.37E−02 (Dropped) (Dropped) (Dropped) −5.78E+ 00 −5.54E+ 00 −5.13E+ 00 −4.63E+ 00 −4.29E+ 00 −3.60E+ 00

793

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 08 ) 78 7 –7 97

Table 2 (continued)

cut7 cut8 cut9 cut10 No. of observations Log-likelihood Pseudo-R2 F-test (P N F) c LR-test (P N χ2) d

All a

West a

East a

2004

1999

1994

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

− 1.71E+ 00 − 1.02E+ 00 3.77E− 02 7.13E− 01 23,014 − 40,803 0.0821 0.0000

−2.17E+ 00 −1.49E+ 00 −4.39E−01 2.40E−01 17,437 −30,760 0.0775 0.0000

−2.28E+ 00 −1.57E+ 00 −4.52E− 01 2.20E−01 5577 −9949 0.0777 0.0000

−2.26E+ 00 −1.57E+ 00 −4.88E− 01 2.09E−01 10,396 −18,416 0.0851 0.0000 0.0000

−2.58E+ 00 −1.90E+ 00 −8.33E− 01 −1.11E− 01 6570 −115,404 0.0807 0.0000 0.0000

−3.16E+ 00 −2.47E+ 00 −1.44E+ 00 −8.33E− 01 6048 −10,741 0.0850 0.0000 0.0000

Note: Significance at the ten-percent level is indicated by ⁎, significance at the five-percent level is indicated by ⁎⁎ and significance at the onepercent level is indicated by ⁎⁎⁎. a These have been adjusted for heteroscedasticity and clustering at the level of the household. b In our analysis, the results are formatted like a single model when, in fact, there are ten equations in the model because there are eleven levels of subjective well-being. In orderd probit regression, Stata sets the constant to zero and estimates the cut points for separating the various levels of the response variable. Other programs may parameterise the model differently by estimating the constant and setting the first cut point to zero. c F-test on joint significance of AIR and NOISE. d A Likelihood ratio test is not applicable if standard errors have been adjusted for heteroscedasticity.

able”) costs about 2.3% of all household income for air quality improvements and 1.3% for noise.

5.2.

The hedonic model

Turning now to the hedonic analysis reported in Table 3, it is apparent that perceived noise and air pollution are not generally statistically significant. This conclusion is unaltered if data from different years is analysed separately, or if the data are analysed according to whether the household is situated in former East or West Germany.6 Elsewhere we find that the absence of central heating and poor renovation reduce the logged price per square meter. There are also marked differences according to the type of dwelling (flat, bungalow etc.) and according to the size of the community with more expensive properties being located in major urban areas. The rent per square meter varies with the number of square meters occupied suggesting that significant transaction costs prevent owners from dividing larger dwellings into separate apartments. The average rate of unemployment of a Federal State is negative and significant in the specifications for 1994 and 1999. The coefficient is positive in the aggregated models. Higher average levels of GDP per capita at State level seem to increase house prices in some specifications. In the regression including only Western Germany or data for 1994 the RESET tests for functional form points to a possible misspecifications. The evidence therefore suggests that even though perceived air quality and noise levels contribute to individual welfare, differences in the perceived levels of these environmental variables are not capitalised into house prices. At least in Germany using the hedonic technique to estimate the value of air quality and noise nuisance risks underestimating the extent to which they contribute to SWB. 6

Note that the inclusion of air and noise quality as categorical rather than continuous variables does not result in a statistically significant results.

6.

Discussion

The current results require us to explain why perceived air quality and noise levels contribute to SWB but are statistically insignificant determinants of house prices, particularly when many other hedonic house price studies have uncovered significant parameter values. One possibility is that hedonic studies have generally considered much smaller geographical areas, typically cities rather than an entire country. It may be that housing markets in Germany are geographically segmented and that assuming a national market obscures the impact that these variables have on house prices. But defining geographically smaller areas such as Hamburg, Bremen or Berlin does not increase the statistical significance of perceived air quality and noise nuisance and neither does distinguishing between communities of differing sizes (results not shown). Our explanation is that the majority of hedonic studies appear to be undertaken in the United States. As pointed out in Section 3, in Germany by contrast people are less mobile. Also, the number of individuals renting homes is much greater and the rental sector is subject to regulations governing when and under what circumstances prices can be increased or tenants be evicted (see e.g. Grundmann, 2001, for some recent developments on the regulation of rental housing). These might explain why the price of housing does not appear to reflect variations in the level of environmental attributes. Similarly, Van Praag and Baarsma (2005) found no significant relationship between house prices and noise levels for households located around the Amsterdam airport. They explained their findings by the peculiarities of the Dutch property market which result in relatively large switching costs as well as the non-monetary costs of moving. If we are correct then techniques based on SWB might prove to be a useful addition to valuation techniques to reveal the total shadow costs of environmental qualities in contexts where the underlying assumptions of hedonic technique appear unreasonable.

794

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 8 ) 7 8 7 –7 97

Table 3 – Hedonic house prices, noise and air quality All a

Owner a

Renter a

West a

East a

1999

1994

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

Coefficient

6.95E−01 ⁎⁎ 3.66E−03 −1.10E− 03 2.50E−01 ⁎⁎⁎ −2.64E− 02⁎⁎⁎ −1.42E− 01⁎⁎⁎

2.04E+ 00 ⁎⁎⁎ 2.31E−03 −5.27E−03 (Dropped) (Dropped) −1.07E−01 ⁎⁎⁎ 5.32E−02 ⁎⁎ 1.12E−01 ⁎⁎⁎ 3.44E−03 ⁎⁎⁎ −3.56E−02 ⁎⁎ −4.38E−02 ⁎

9.52E− 01 ⁎⁎⁎ 3.39E− 03 1.66E− 03 (Dropped) 4.96E− 03 − 1.45E−01 ⁎⁎⁎

1.87E+ 00 ⁎⁎⁎ 3.97E−03 −2.36E− 03 2.07E−01 ⁎⁎⁎ −9.30E− 02 ⁎⁎⁎ −1.72E− 01 ⁎⁎⁎

−2.22E+ 00 1.17E−02 3.67E−03 2.61E−01 ⁎⁎⁎

2.11E+ 00 ⁎⁎⁎ − 3.21E−03 − 2.70E−04 1.83E− 01 ⁎⁎⁎ − 1.05E−01 ⁎⁎⁎ − 1.39E−01 ⁎⁎⁎

− 2.52E−03 4.14E− 02 ⁎⁎⁎ 3.98E− 03 ⁎⁎⁎ − 6.54E−02 ⁎⁎⁎ 3.95E− 02 ⁎⁎

2.04E−02 4.65E−02 −1.55E−03 1.11E−01 −5.64E−02 1.70E−01 2.35E−01 2.62E−01 1.42E−01 1.71E−01 2.11E−01 2.27E−01 2.83E−02 ⁎⁎⁎ −3.10E−03 ⁎⁎⁎ −2.03E−02 ⁎⁎⁎

− 8.05E−02 − 9.54E−02 − 7.36E−02 − 3.66E−02 − 9.41E−02 1.23E− 01 1.77E− 01 2.02E− 01 2.17E− 01 2.08E− 01 2.02E− 01 2.23E− 01 3.31E− 02 ⁎⁎⁎ − 3.33E−03 ⁎⁎⁎ − 1.16E−02 ⁎⁎⁎ − 2.21E−02 ⁎⁎ − 1.17E−01 ⁎⁎⁎ − 1.25E−01 ⁎⁎⁎ − 1.16E−01 ⁎⁎⁎ − 1.05E−01 ⁎⁎⁎ 3.13E− 05⁎⁎⁎ 2.81E− 02 ⁎⁎⁎ 3.62E− 01 ⁎⁎⁎

1.29E−03 6.79E−02 ⁎⁎⁎ 4.45E−03 ⁎⁎⁎ −3.19E− 02 ⁎⁎⁎ −3.86E− 03 −4.52E− 02 −1.70E− 02 −3.39E− 02 5.47E−03 −7.13E− 02 1.73E−01 2.28E−01 ⁎⁎ 2.74E−01 ⁎⁎ 2.25E−01 ⁎⁎ 2.19E−01 ⁎⁎ 2.35E−01 ⁎⁎ 2.39E−01 ⁎⁎ 3.43E−02 ⁎⁎⁎ −3.08E− 03 ⁎⁎⁎ −1.50E− 02 ⁎⁎⁎ −2.29E− 02 ⁎⁎ −3.40E− 01 ⁎⁎⁎ −2.29E− 01 ⁎⁎⁎ −1.72E− 01 ⁎⁎⁎ −1.55E− 01 ⁎⁎⁎ 2.90E−02 ⁎⁎⁎

1.88E+ 00 ⁎⁎⁎ 1.23E−02 ⁎ 1.42E−03 4.33E−01 ⁎⁎⁎ 1.09E−01 ⁎⁎⁎ −1.05E− 01 ⁎⁎⁎ 1.45E−02 5.89E−02 ⁎⁎⁎ 4.48E−03 ⁎⁎⁎ −4.66E− 02 ⁎⁎⁎

CONSTANT AIR NOISE OWNER SOCIAL NO-HEATING GARDEN BALCONY MOVEIN CONDITION WINDOW AREA1 AREA2 AREA3 AREA4 AREA5 TYPE1 TYPE2 TYPE3 TYPE4 TYPE5 TYPE6 TYPE7 BUILT SQUARE CITY TRANSPORT GGK1 GGK2 GGK3 GGK4 UNEMPL_STATE GDP_STATE STATE1 STATE2 STATE3 STATE4 STATE5 STATE6 STATE7 STATE8 STATE9 STATE10 STATE11 STATE12 STATE13 STATE14 STATE15 EAST 1999

8.93E−02 1.08E−01 ⁎⁎⁎ 9.86E−02 ⁎⁎⁎ 7.76E−02 ⁎⁎ (Dropped) 1.29E−01 ⁎⁎⁎ 1.62E−01 4.75E−03

1.30E−05 8.35E−02 −1.55E−01 −1.53E−01 −5.55E−01 −1.07E−01 −5.62E−02 −1.10E−01 −2.73E−02 −9.52E−02 −1.88E−01 −1.34E−01 (Dropped) −1.70E−01 −2.33E−01 −1.44E−01 (Dropped) −1.05E−02

R2 F-test (P N F) b RESET–test (P N F) c

0.3481 0.7035 0.7120

0.3199 0.8061 0.7610

2.77E−03 5.48E−02⁎⁎⁎ 4.01E−03 ⁎⁎⁎ −6.81E− 02 ⁎⁎⁎ 1.32E−02 −1.50E− 02 −1.66E− 02 −1.79E−02 1.80E−02 −5.73E− 02 1.72E−01 2.49E−01 ⁎⁎ 2.80E−01 ⁎⁎ 2.44E−01 ⁎⁎ 2.34E−01 ⁎⁎ 2.19E−01 ⁎ 2.29E−01 ⁎⁎ 3.14E−02 ⁎⁎⁎ −3.30E− 03 ⁎⁎⁎ −1.51E− 02 ⁎⁎⁎ −1.75E− 02 ⁎⁎ −2.22E− 01 ⁎⁎⁎ −1.80E− 01 ⁎⁎⁎ −1.54E− 01 ⁎⁎⁎ −1.31E− 01 ⁎⁎⁎ 3.52E−02 ⁎⁎⁎ 2.84E−05 ⁎⁎⁎ 6.01E−01 ⁎⁎⁎ (Dropped) 4.04E−01 ⁎⁎⁎ −1.92E− 02 3.83E−01 ⁎⁎⁎ 4.11E−01 ⁎⁎⁎ 4.70E−01 ⁎⁎⁎ 4.87E−01 ⁎⁎⁎ 4.51E−01 ⁎⁎⁎

−1.38E−02 −4.35E−01 −3.44E−01 −2.84E−01 −2.36E−01 1.83E−02 ⁎

⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎

⁎⁎

⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎

− 2.86E−01 2.09E− 01 ⁎⁎⁎ − 1.85E−01 1.41E− 01 ⁎

−2.05E− 06 1.45E−01 ⁎⁎⁎ 3.32E−02 −7.41E− 02 ⁎⁎⁎ −2.34E− 01 ⁎⁎ −3.71E− 02 1.18E−01⁎ (Dropped) 1.34E−01 ⁎⁎⁎ 1.10E−01 ⁎⁎ −3.89E− 01 ⁎⁎⁎

−2.14E−02 −8.51E−02 ⁎⁎⁎ −9.27E−03 4.27E−02 ⁎⁎ 1.64E−03 ⁎⁎ −1.08E−01 ⁎⁎⁎ 2.61E−02 6.62E−02 ⁎ 7.28E−03 3.09E−02 1.49E−01 (Dropped) (Dropped) 1.48E−01 ⁎⁎ 8.96E−02 1.16E−01 1.06E−01 6.06E−02 8.83E−02 4.08E−02 ⁎⁎⁎ −4.06E−03 ⁎⁎⁎ −1.41E−02 ⁎⁎ 1.51E−02 −5.13E−03 3.55E−02 9.15E−03 2.58E−02 −7.92E−03 3.13E−04 (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) 8.65E−02 ⁎⁎

− 1.88E−03 5.27E− 02 ⁎⁎⁎ 3.06E−03 ⁎⁎⁎ − 6.61E−02 ⁎⁎⁎ − 1.33E−03 − 7.83E−02 − 6.33E−02 − 6.96E−02 − 4.48E−02 − 1.35E−01 3.63E− 01 ⁎ 4.37E− 01 ⁎⁎ 4.68E− 01 ⁎⁎ 4.37E− 01 ⁎⁎ 4.25E− 01 ⁎⁎ 4.24E− 01 ⁎⁎ 4.36E− 01 ⁎⁎ 3.08E− 02 ⁎⁎⁎ − 2.91E−03 ⁎⁎⁎ − 1.58E−02 ⁎⁎⁎ − 1.42E−02 − 2.86E−01 ⁎⁎⁎ − 2.44E−01 ⁎⁎⁎ − 2.04E−01 ⁎⁎⁎ − 1.88E−01 ⁎⁎⁎ − 5.01E−03 ⁎⁎ 4.01E− 06 ⁎⁎ 1.77E− 01 ⁎⁎⁎ (Dropped) − 4.45E−03 − 1.62E−01 ⁎⁎⁎ 2.94E− 02 ⁎ 6.73E− 02 ⁎⁎⁎

2.23E−02 3.85E−02 2.21E−02 1.88E−02 9.48E−02 1.69E−02 2.72E−02 1.08E−01 1.44E−01 1.10E−01 1.07E−01 7.67E−02 1.07E−01 4.41E−02 ⁎⁎⁎ −3.63E− 03 ⁎⁎⁎ −1.34E− 02 ⁎⁎⁎ −2.39E− 02 ⁎⁎ −2.14E− 01 ⁎⁎⁎ −1.48E− 01 ⁎⁎⁎ −1.25E− 01 ⁎⁎⁎ −1.16E− 01 ⁎⁎⁎ −1.21E− 02 ⁎⁎ 5.39E−06 ⁎⁎ 1.54E−01 ⁎⁎⁎ (Dropped) 3.21E−02 −2.57E− 02 3.47E−02 7.07E−02 ⁎⁎ (Dropped) 4.46E−02 ⁎

1.55E− 01 2.72E− 01 ⁎⁎⁎ 2.26E− 01 ⁎⁎ 2.11E− 01 ⁎⁎ − 1.06E−01 7.40E− 02 ⁎⁎ 4.04E− 02 5.16E− 02 ⁎ (Dropped) 1.16E− 01 ⁎⁎⁎ (Dropped) 6.26E− 02 ⁎⁎

(Dropped) (Dropped) (Dropped) (Dropped) (Dropped) (Dropped) 1.62E−02

−5.84E−02 ⁎ 1.85E−01 1.64E−01 (Dropped) −6.40E−01

1.03E− 02 (Dropped) 4.67E− 02 ⁎⁎ (Dropped) (Dropped)

1.32E−02 −6.40E− 02 ⁎⁎ −2.53E− 02 −2.61E− 03 −2.80E− 02 (Dropped) −2.43E− 02 (Dropped) (Dropped)

0.3796 0.5115 0.3750

0.6064 0.7700 0.0240

0.4127 0.1210 0.3370

0.2862 0.7378 0.3210

0.3661 0.0359 0.0000

(Dropped) 3.65E− 02 5.58E− 03 − 1.15E−01 ⁎⁎⁎ 5.24E− 02 ⁎⁎ 1.16E− 01 ⁎⁎⁎

Note: Significance at the ten-percent level is indicated by ⁎, significance at the five-percent level is indicated by ⁎⁎ and significance at the onepercent level is indicated by ⁎⁎⁎. a These have been adjusted for heteroscedasticity and clustering at the level of the household. b F-test on joint significance of AIR and NOISE. c Ramsey RESET test for functional form.

Finally the reader should be aware that variations in selfperceived levels of air pollution and noise nuisance might themselves be correlated with other unobserved character-

istics of individuals. Thus unhappy or depressed individuals might perceive noise nuisance or poor air quality more acutely than others. Nonetheless individuals living close to the centre

795

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 08 ) 78 7 –7 97

Table 4 – Variation in air quality Affected by air pollution on residential area 1 (not at all) 2 3 4 5 (very strongly)

Living close to the city centre

Living in rural area

Living close to transport link

No transport link in walking distance

Living in industrial area

Living in residential area

In %

In %

In %

In %

In %

In %

28.62 39.57 21.68 7.39 2.74

46.17 38.74 9.98 4.19 0.92

37.61 40.58 14.82 5.31 1.68

58.17 32.69 5.77 2.40 0.96

20.41 42.86 20.41 11.56 4.76

46.24 39.71 10.31 2.89 0.85

Source: German socio-economic panel, own calculation.

of big cities, close to transport links and in mixed-use or predominantly industrial areas are far more likely to declare themselves strongly affected than individuals living in other locations (see Tables 4 and 5). Thus individual assessments of environmental quality are not wholly determined by individual traits. There is also the possibility that some individuals might exaggerate noise and air pollution if they felt that would encourage the authorities to tackle these problems although neither air pollution nor noise nuisance is the focus of the SOEP. Although it would be surprising to find no significant relationship between the two subjective measures (well-being and perceived air and noise pollution) as feeling affected by pollution is equivalent to saying one is unhappy about pollution, it is worth noting that both measures are consistent with each other. Of course, objective measures of air pollution and noise nuisance would have been preferred to selfperceived levels. However, local measures are generally unavailable as there are only a limited number of monitoring stations per town or region. Interpolating pollution measures would create errors of its own. Neither is it clear which of a variety of metrics (daily average, peak, night time etc.) is more relevant. Interestingly, Van Praag and Baarsma (2005) who had information on both subjective and objective noise levels found no statistically significant relationship between SWB

and objective noise nuisance. In their analysis they continue with replacing the objectively measured noise level by a subjective translation which results in a significant coefficient for noise. Unlike in our study they are then able to compute the monetary value of a change in objective noise since they possess data on both objective and subjective noise levels.

7.

Conclusion

This paper compares the extent to which perceived measures or air quality and noise nuisance are reflected in measures of SWB and property prices. At least in the context of Germany it appears that perceived levels of air pollution and noise nuisance are not capitalised into property prices. This is most likely due to the fact that the housing market is highly regulated by the state. By contrast high noise levels and poor air quality markedly diminish SWB. Future analyses might attempt to apply the same techniques in other countries specifically those in which the hedonic technique has been observed to yield significant results for air pollution and noise nuisance. These analyses might also be able to compare selfperceived measures of air pollution and noise nuisance with their objectively measured counterparts. Further research might also attempt to determine whether individuals sharing the same characteristics reach the same level of well-being

Table 5 – Variation in noise nuisance Affected by noise pollution on residential area 1 (not at all) 2 3 4 5 (very strongly)

Living close to the city centre

Living in rural area

Living close to transport link

No transport link in walking distance

Living in industrial area

Living in residential area

In %

In %

In %

In %

In %

In %

29.28 39.58 20.71 7.19 3.24

44.26 36.03 12.99 4.84 1.88

35.53 39.45 16.35 6.38 2.29

57.42 31.58 8.13 1.44 1.44

23.81 39.46 24.49 8.16 4.08

44.53 38.83 11.84 3.64 1.16

Source: German socio-economic panel, own calculation.

796

EC O LO GIC A L E CO N O M ICS 6 4 ( 2 00 8 ) 7 8 7 –7 97

irrespective of their geographical location since this is the key assumption underpinning the hedonic technique.

This paper benefits greatly from comments by Richard S.J. Tol and two anonymous referees. The Michael Otto Foundation for Environmental Protection provided welcome financial support. A version of the paper was presented at the 7th Conference of the International Society for Quality-of-Life Studies. All errors and opinions are ours.

Appendix A Table A1. How strongly do you feel you are affected by air pollution in your residential area? All

West

East

2004

1999

1994

In %

In %

In %

In %

In %

In %

38.44 40.61 14.22 5.18 1.55

41.11 39.20 13.61 4.69 1.39

30.12 45.01 16.14 6.69 2.04

47.32 39.15 9.76 3.11 0.67

35.59 42.69 15.02 5.12 1.58

26.46 40.84 20.94 8.76 3.00

Source: German socio-economic panel, own calculation.

Table A2. How strongly do you feel you are affected by noise pollution in your residential area?

Not at all Slightly Bearable Strongly Very strongly

All

West

East

2004

1999

1994

In %

In %

In %

In %

In %

In %

36.57 39.25 15.93 6.10 2.16

39.38 38.14 15.12 5.61 1.76

27.82 42.72 18.45 7.63 3.39

42.39 38.02 13.09 4.97 1.53

33.69 42.24 16.36 5.68 2.03

29.80 38.10 20.27 8.47 3.36

Source: German socio-economic panel, own calculation.

Table A3. How satisfied are you with your life, all things considered?

0 1 2 3 4 5 6 7 8 9 10

Model

All

West

East

2004

1999

1994

In %

In %

In %

In %

In %

In %

0.56 0.44 1.45 3.42 4.30 13.31 12.37 22.67 28.61 8.68 4.19

0.56 0.39 1.29 2.99 3.81 11.74 11.40 22.44 30.42 9.98 4.99

0.59 0.59 1.95 4.75 5.82 18.20 15.40 23.38 22.97 4.65 1.7

0.58 0.45 1.56 3.67 4.70 13.79 12.50 22.66 28.22 8.28 3.61

0.53 0.42 1.29 3.01 3.91 12.40 12.19 22.53 29.85 9.63 4.21

0.59 0.44 1.44 3.44 4.05 13.47 12.35 22.83 27.92 8.33 5.14

“0” means completely dissatisfied, “10” means completely satisfied? Source: German socio-economic panel, own calculation.

Air

Noise

Model

−5.07E −02⁎⁎⁎ −8.52E −02⁎⁎⁎

−4.82E −02⁎⁎⁎ 4.21E −03

2004

Air

Noise

−5.21E −02⁎⁎⁎ EAST 1999 −6.49E −02⁎⁎⁎ 1994 −7.41E −02⁎⁎⁎ χ2(1) = Parameter χ2(2) = Parameter χ2(1) = ⁎ 3.29 homogeneity 0.74 homogeneity 1.68 test test Variance −0.060⁎⁎⁎ −0.038⁎⁎⁎ Variance −0.063⁎⁎⁎ weighted weighted estimate estimate WEST

Acknowledgements

Not at all Slightly Bearable Strongly Very strongly

Table A4. Parameter homogeneity among the coefficients

− 6.53E − 02⁎⁎⁎ − 1.37E − 02 − 1.16E − 02 χ2(2) = 6.91⁎⁎ − 0.036⁎⁎⁎

Note: Significance at the ten-percent level is indicated by ⁎, significance at the five-percent level is indicated by ⁎⁎ and significance at the one-percent level is indicated by ⁎⁎⁎.

Appendix B To be able to calculate changes in levels of SWB by changes in explanatory variables, we start with deriving the probabilities for an individual to choose one of the 11 scores on lifesatisfaction. Using the information on the cutpoints reported in Table 2, the chance that a respondent for being least satisfied (score 0) is Prob (“score 0”) = F (cut1 − H), where F is the standard normal distribution and H is a person's SWB (which is the predicted value of the fitted model from the ordered probit regression). Consequently, the probability to score 1 is Prob (“score 1”) = F (cut2 − H) − F (cut1 − H). Probabilities for other levels of SWB can be computed alike (for more details see Greene, 2003). Table B1 shows the probabilities derived from our model estimates. They are summarized in steps of three (scores 0 to 2, scores 4 to 6 and score 8 to 10) for all our model specifications. It is evident that the probability to score at the bottom of the scale is very low; mainly less than 5%. Scores of 5 to 8 (not shown) are dominating. This corresponds to the original data (compare Table A3 in Appendix A).

Table B1: Probability for a person to report a certain score of SWB Score of lifesatisfaction

All

West

East

2004

1999

1994

In %

In %

In %

In %

In %

In %

0 (lowest) to 2 4 to 6 8 to 10 (highest)

4.63 48.92 16.98

1.32 32.36 36.95

1.35 42.84 25.78

1.08 33.81 35.57

0.70 27.46 43.90

0.13 14.07 63.87

Source: Own calculation.

Having computed the probabilities to choose one of the 11 scores on life-satisfaction makes it possible to derive the impact of marginal changes of explanatory variables on changes in a person's SWB. Taking the example from above, they can be derived as follows. The change in ones probability of being, for example, least satisfied is ΔProb (“score 0”) = F (cut1 - (H + ΔH)) − F (cut1 − H). Table B2 shows the calculated

EC O L O G IC A L E C O N O M IC S 6 4 ( 2 0 08 ) 78 7 –7 97

changes in probabilities when the quality of air increases or noise exposure decreases. Changes are, again, summarized in steps of three for all our model specifications. In our model, improvements in environmental services imply that respondents move from their current category to the next higher (e.g. from “very strongly affected” to “strongly affected”). If the quality of environmental services increases the distribution of SWB changes. The probability for a respondent to score 8 or higher increases by 2 to 3% for changes in air pollution and by 1 to 2% for improvements in noise exposure. The representation of all other categories diminishes. As a consequence of the estimated coefficients for air and noise, changes in air quality have more pronounced effects although more people feel negatively effected by noise pollution (compare Tables A1 and A2 in Appendix A). TableB2: B2:Change Change in in probability probability when Table when environmental environmental quality changes quality changes All

West

East

2004

1999

1994

In %

In %

In %

In %

In %

In %

Quality of air increases: 0 (lowest) to 2 − 0.62 4 to 6 − 1.09 8 to 10 (highest) 1.50

−0.18 −1.46 1.90

− 0.32 − 2.49 2.67

− 0.16 − 1.55 1.92

−0.14 −1.91 2.55

−0.04 −1.63 2.81

Noise exposure decreases: 0 (lowest) to 2 − 0.34 4 to 6 − 0.63 8 to 10 (highest) 0.85

−0.17 −1.39 1.80

0.011 0.131 − 0.141

− 0.20 − 1.94 2.40

−0.031 −0.401 0.541

−0.01 −0.25 0.44

1

Results are statistically not significant. Source: Own calculation.

Similarly, the impact of changes in other explanatory variables like income on SWB can be computed. An increase in household's net income per month of e.g. EUR 100 increases the probability of a respondent to score 8 or higher by only 0.36%.

REFERENCES Bateman, I.T., Carson, R.T., Day, B., Hanemann, M., Hanley, N., Hett, T., Jones-Lee, M., Loomes, G., Mourato, S., Özdemiroglu, E., Pearce, D.W., Sugden, R., Swanson, J., 2002. Economic Valuation with Stated Preference Techniques: a Manual. Edward Elgar, Cheltenham. Bundesagentur für Arbeit, 2006. Arbeitsmarktstatistiken der Bundesagentur für Arbeit. Bundesagentur für Arbeit, Nürnberg. Clark, A.E., Oswald, A.J., 1994. Unhappiness and unemployment. Economic Journal 104, 648–659. Di Tella, R., MacCulloch, R.J., Oswald, A.J., 2001. Preferences over inflation and unemployment: evidence from surveys of happiness. The American Economic Review 91, 335–341. Di Tella, R., MacCulloch, R.J., Oswald, A.J., 2003. The macroeconomics of happiness. Review of Economics and Statistics 85, 809–827. Easterlin, R.A., 2001. Income and happiness: towards a unified theory. Economic Journal 111, 465–484. Ferrer-i-Carbonell, A., Gowdy, J.M., 2007. Environmental degradation and happiness. Ecological Economics 60 (3), 509–516. Freeman III, A.M., 2003. The measurement of environmental resource values. Resources for the Future. Washington, D.C.

797

Frey, B.S., Stutzer, A., 2002. Happiness and Economics. Princeton University Press, Princeton. Frijters, P., Van Praag, B.M.S., 1998. The effects of climate on welfare and well-being in Russia. Climatic Change 39, 61–81. Greene, W.H., 2003. Econometric Analysis, 5th edition. Prentice Hall, New York. Grundmann, B., 2001. Die Mietrechtsreform. Neue Juristische Wochenschrift, pp. 2498–2505. Hoffmann, J., Kurz, C., 2002. Rent indices for housing in West Germany 1985 to 1998. Discussion Paper 01/02. Economic Research Centre of the Deutsche Bundesbank, Frankfurt. Holm-Müller, K., Hansen, H., Klockmann, M., Luther, P., 1991. Die Nachfrage nach Umweltqualität in der Bundesrepublik Deutschland. Forschungsbericht 10103110/11. Umweltbundesamt, Berlin. Kahnemann, D., Sudgen, R., 2005. Experienced utility as a standard of policy evaluation. Environmental and Resource Economics 32, 161–181. Kahnemann, D., Wakker, P.P., Sarin, R., 1997. Back to Bentham? Explorations of experienced utility. The Quarterly Journal of Economics 112 (2), 375–405. Mitchell, R.C., Carson, R.T., 1989. Using Surveys to Value Public Goods. Resources for the Future. Washington, D.C. Nelson, J.P., 2004. Meta-analysis of airport noise and hedonic property values: problems and prospects. Journal of Transport Economics and Policy 38 (1), 1–27 January. Rehdanz, K., Maddison, D., 2005. Climate and happiness Ecological Economics 52, 111–125. Roback, J., 1982. Wages, rents, and the quality of life. Journal of Political Economy 90, 1257–1278. Rosen, S., 1974. Hedonic prices and implicit markets: product differentiation in pure competition. Journal of Political Economy 82, 34–55. Schulz, W., 1985. Der monetäre Wert besserer Luft: Eine empirische Analyse individueller Zahlungsbereitschaften und ihrer Determinanten auf der Basis von Repräsentativumfragen. Verlag Peter Lang, Berlin. Smith, V.K., Huang, J.C., 1995. Can markets value air quality? A meta-analysis of hedonic property value models. Journal of Political Economy 103, 209–227. Statistisches Bundesamt, 2006a. Verbraucherpreisindex und Index der Einzelhandelspreise: Jahresdurchschnitte ab 1948. Statistisches Bundesamt, Wiesbaden. Statistisches Bundesamt, 2006b. Volkswirtschaftliche Gesamtrechnungen der Länder, Bruttoinlandsprodukt, Bruttowertschöpfung in den Ländern und Ost-West-Groβraumregionen Deutschlands 1991 bis 2005. Statistisches Bundesamt, Wiesbaden. Umweltbundesamt, 2005. Hintergrundpapier zum Thema Staub/Feinstaub (PM). Umweltbundesamt, Berlin. Umweltbundesamt, 2006. Luftreinhaltestategien Umweltbundesamt, Berlin. (published 22.02.2006). http://www. umweltbundesamt.de/luft/reinhaltestrategien/index.htm. Van Praag, B.M.S., Baarsma, B.E., 2005. Using happiness surveys to value intangibles: the case of airport noise. Economic Journal 115, 224–246. Weinberger, M., Thomassen, H.G., Willeke, R., 1991. Kosten des Lärms in der Bundesrepublik Deutschland: Forschungsbericht 101 03 110/05. Umweltforschungsplan des Bundesministers für Umwelt, Naturschutz und Reaktorsicherheit: Umweltplanung, Ökologie. In: Umweltbundesamt (Ed.), Erich Schmidt Verlag, Berlin. Welsch, H., 2002. Preferences over prosperity and pollution: environmental valuation based on happiness surveys. Kyklos 55, 473–494. Welsch, H., 2006. Environment and happiness: valuation of air pollution using life-satisfaction data. Ecological Economics 58, 801–813.